Across universities worldwide, students and educators are navigating one of the most pressing challenges in modern academia: how to fairly and accurately assess work in an era of widely accessible artificial intelligence. For many institutions, AI detection tools seemed like a straightforward answer — but Nanyang Technological University (NTU) in Singapore has reached a starkly different conclusion. NTU’s decision to phase out AI plagiarism detectors entirely signals a major shift in how higher education institutions are rethinking academic integrity in the age of generative AI.
NTU Phases Out Its Institutional AI Detector, Citing Fundamental Unreliability
NTU will discontinue the use of its institutional AI detector from 2027, with the tool scheduled for deactivation at the end of the current semester on December 31. The decision was communicated to faculty members in an email signed by associate vice-provost for education transformation Tan Seng Chee, as reported by Singapore-based outlet CNA.
Tan stated plainly that the tools lack “empirical validity” and are “fundamentally unreliable” for the purposes of proving academic misconduct. More critically, the university determined that AI probability scores cannot distinguish between unauthorised AI use and responsible, curriculum-sanctioned use of the same technology. Faculty members have been explicitly instructed that automated AI scores or detection flags must not be presented as evidence of academic violations.
AI Detection Tools Have Created Real Problems for Real Students
The practical harm caused by AI detectors has not been theoretical. NTU’s own internal guidance acknowledges that these tools are prone to producing both false positives and false negatives, meaning innocent students can be flagged and genuine misuse can go undetected. The technology can also be bypassed through minor textual edits, which undermines any claim to robustness as an enforcement mechanism.
Of particular concern is the documented risk that AI detectors disproportionately disadvantage students whose writing reflects non-native language patterns. In a diverse academic community, this creates a structural fairness problem — one that punishes students not for cheating, but for writing in a style that detection algorithms misread as machine-generated. Ambiguous detection reports also place unnecessary administrative burdens on faculty and erode trust between teachers and students, according to NTU’s communication to staff.
The university’s concern about real-world impact is not without precedent. A 2025 controversy involving three NTU students accused of AI misuse — who disputed the charges and raised due process concerns — brought the limitations of automated detection into sharp institutional focus.
NTU Moves Toward Assessment Methods That Respect Student Learning Processes
Rather than relying on surveillance-based detection, NTU is shifting toward assessment approaches that document how students develop and complete their work. This includes standardised disclosures of AI use, which allows the university to evaluate transparency and intentionality rather than attempting to forensically identify AI involvement after the fact.
This approach reflects a growing consensus among educators that the question is no longer whether students use AI, but how they use it and whether they understand what they are doing. By focusing on process documentation, NTU positions assessment as a tool for learning accountability rather than a mechanism for catching rule-breakers.
The shift also aligns with NTU’s broader expansion of access to institutional AI tools for both faculty and students. University leadership noted that continuing to operate automated detection systems alongside actively sanctioned AI use would send contradictory signals to the academic community.
A Broader Debate in Higher Education About Whether AI Detectors Can Ever Be Trusted
NTU’s decision reflects a debate that extends well beyond a single institution. Universities globally have grappled with whether any AI detection technology can meet the evidentiary standard required to sanction a student for cheating. The core technical problem is that AI-generated text and human-written text increasingly overlap in ways that no current tool can reliably parse — and the stakes of getting that assessment wrong, for a student’s academic record and future, are significant.
By taking the definitive step of deactivating its detector and establishing a clear policy that AI scores cannot serve as evidence of misconduct, NTU has staked out one of the most explicit institutional positions on this question seen in the region.
Here’s What You Need to Know About NTU’s AI Detector Decision
When exactly will NTU stop using its AI detector? NTU will deactivate its institutional AI detector on December 31, marking the end of the current semester, and will formally discontinue its use from 2027 onward.
Why has NTU decided to stop using AI plagiarism detection tools? NTU stated that AI detection tools are “fundamentally unreliable” and lack “empirical validity,” meaning they cannot accurately or fairly prove that a student has used AI in an unauthorised way.
Can AI detection scores still be used as evidence of cheating at NTU? No. NTU has explicitly instructed faculty members that automated AI probability scores and detection flags must not be presented as evidence of academic violations under any circumstances.
Do AI detectors produce errors that could harm innocent students? Yes. NTU’s own guidance confirms that AI detectors produce both false positives and false negatives, can be circumvented through minor text edits, and may unfairly disadvantage students who write in non-native English patterns.
What will NTU use instead of AI detectors to assess academic integrity? NTU will move toward assessment methods that document how students develop and complete their work over time, including standardised disclosures of AI use, replacing automated surveillance with process-based evaluation.
Did a specific incident contribute to NTU’s decision to phase out AI detectors? A 2025 controversy involving three NTU students accused of AI misuse, who disputed the charges and raised due process concerns, contributed to the university’s reassessment of how AI detection tools are applied in academic misconduct proceedings.
Does NTU support students using AI tools for their studies? Yes. NTU has been expanding access to institutional AI tools for both faculty and students, and leadership noted that continuing to use AI detectors alongside sanctioned AI use would send contradictory and confusing signals to the university community.
NTU’s move to eliminate AI plagiarism detectors from its academic integrity processes marks a significant moment for higher education in the region. The university’s conclusion — that these tools are too unreliable to use as evidence against students — challenges institutions still relying on automated detection to reconsider whether the technology can ever meet the standard of fairness that students deserve.
